Development of the Exercise Motives and Gains Inventory

There are existing measures of exercise motives (what people want from exercise), but corresponding measures of gains (what people get) are needed, because motives and gains could influence each other and together influence other variables. An exercise motives and gains inventory (EMGI) was developed by creating gains scales to complement existing Exercise Motivations Inventory 2 scales. Confirmatory factor analyses of EMGI items established that items reflected their intended constructs; and that motive and gain constructs were distinct. Exploratory structural equation modeling of EMGI scales established that the higher-order structures of motives and gains were somewhat different: Appearance motive was associated with weight management, whereas appearance gain was associated with health and fitness. Paired-sample t-tests established that gains were less than motives in some instances (ill-health avoidance, positive health), and greater in others (e.g., affiliation, challenge). The EMGI can be used to investigate the consequences and causes of motives and gains.


INTRODUCTION Aim
Participatory motives are what individuals seek to attain or avoid by engaging in a particular domain of behavior. The study of such motives has become an important cornerstone of exercise participation research (Ingledew & Markland, 2008). However, whereas motives have received ample attention, gains have not. By gains, the authors mean what people have attained or avoided through engagement. Arguably, motives (what people want) and gains (what they get) should be studied in parallel, because they are likely to influence each other and jointly influence exercise-related processes and outcomes such as behavioral regulation, exercise amount, satisfaction, and intention. The aim of the present study was to develop a measure of motives and gains by adding gains scales to an existing measure of motives, the Exercise Motivations Inventory version 2 (EMI-2; Markland & Ingledew, 1997). First, the EMI-2 will be reviewed.
The EMI-2 has demonstrated usefulness in identifying various determinants of exercise motives. Motives have been found to differ by age (Dacey, Baltzell, & Zaichkowsky, 2008;Ingledew & Sullivan, 2002;Quindry et al., 2011); for example, appearance and stress management being lower in older adults rather than in adults not as old (Dacey et al., 2008). Traditional students (full-time, 18-22 years old, living on campus) compared with non-traditional students were more motivated by challenge, social recognition, affiliation, competition, appearance and nimbleness, and less by health pressure and ill-health avoidance (Kulavic et al., 2013). Sex differences have been found (Dacey et al., 2008;Egli et al., 2011;Grogan et al., 2006;Ingledew & Sullivan, 2002;Kilpatrick et al., 2005;Quindry et al., 2011;Shen & Xu, 2008); for example, adolescent females compared with males having higher weight management and lower strength and endurance motives (Ingledew & Sullivan, 2002). Differences by sexual orientation have also been found; for example gay men compared with heterosexual men having higher appearance and lower enjoyment and competition motives (Grogan et al., 2006). Ethnic differences (Egli et al., 2011;Zajac & Schier, 2011) have also been found; for example, Black compared with White students being more motivated by health pressures, ill-health avoidance, and nimbleness and less by stress management, revitalization, enjoyment, and weight management (Egli et al., 2011). Body image has predicted motives (Ingledew & Sullivan, 2002;Zajac & Schier, 2011) differently in males and females. Personality traits have predicted motives; for example, openness positively predicting health/fitness motives and neuroticism positively predicting appearance/weight motives (Ingledew & Markland, 2008). Life goals (what individuals generally aim to attain or avoid in life) have also predicted motives; for example image life goal predicting appearance/weight motives (Ingledew et al., 2009).
The EMI-2 has also demonstrated usefulness in identifying various consequences of exercise motives. Motives have been associated with psychological well-being, for example, appearance motive being associated with poorer well-being in premaintenance exercisers (Maltby & Day, 2001). Motives have predicted behavioral regulation (Ingledew & Markland, 2008;Ingledew et al., 2009); for example, appearance/weight and social recognition predicting external regulation (control by external contingencies), appearance/weight also predicting introjected regulation (control by internalized contingencies), stress management and health/fitness predicting identified regulation (conscious valuing), and affiliation and challenge predicting intrinsic regulation (enjoyment; Ingledew et al., 2009). Through behavioral regulation, motives have predicted amount of exercise participation (Ingledew & Markland, 2008;Ingledew et al., 2009); for example, stress management, health and fitness, affiliation, and challenge positively predicting participation (Ingledew et al., 2009). Motives have also been associated with stage of change (Dacey et al., 2008;Ingledew et al., 1998); for example, appearance and weight motives being prominent in early stages but enjoyment and revitalization motives being conducive to maintenance (Ingledew et al., 1998), with commitment and intention to continue exercising (Funk et al., 2011) and with adherence to an exercise program (Izquierdo-Porrera, Powell, Reiner, & Fontaine, 2002). Finally, motives have been shown to be associated with type of activity (Kilpatrick et al., 2005;Zajac & Schier, 2011); for example, aerobics compared with yoga participants manifesting higher weight management and lower positive health and stress management motives (Zajac & Schier, 2011).
According to Markland and Ingledew (2007), many of these findings can be interpreted in terms of self-determination theory (Ryan & Deci, 2000). From this theoretical perspective, motives lead to autonomous regulation depending upon their potential to satisfy basic needs for autonomy, competence, and relatedness (Vansteenkiste, Lens, & Deci, 2006). Ingledew et al.'s (2009) three-level model of motivation (life goals leading to exercise motives, leading to behavioral regulation, and thereby behavior) is consistent with this theory, although these authors did not measure need satisfaction. However, use of the EMI-2 is not limited to this particular theory and has been used with reference to other frameworks such as Leary and Kowalski's (1990) model of impression management (Strong, Martin Ginis, Mack, & Wilson, 2006). It has also been used in studies without reference to any specific theoretical frameworks (Grogan et al., 2006;Halliwell, Dittmar, & Orsborn, 2007;Izquierdo-Porrera et al., 2002;Kulavic et al., 2013). All in all, if validity is the "degree to which scores on an appropriately administered instrument support inferences about variation in the characteristic that the instrument was 54 developed to measure" (Cizek, 2012), then the cumulative evidence supports the use of the EMI-2 as a measure of exercise motives.

Need for an Exercise Motives and Gains Inventory (EMGI)
Whereas motives are what people seek to attain or avoid through engagement, gains are what they have attained or avoided through engagement. Such a distinction can be found in the research into volunteering behavior where Clary et al. (1998) have studied motives and "functionally relevant benefits" (what we would call gains). Both motives and gains refer in some way to the content (the "what") of behavioral goals (such as engaging in exercise). However, motives are reasons to engage, whereas gains are results from having engaged. Motives and gains are distinct from goal features, such as importance, difficulty, or specificity (see Austin & Vancouver, 1996). For example, two individuals may place the same importance on a behavioral goal, but have different motives for pursuing it and experience different gains from achieving it. When people's gains correspond to their original motives, we would call this motive fulfillment. For example, a person may take up a martial art solely to develop new skills (challenge motive) and find that they do indeed develop such skills (challenge gain). However, unsought gains may occur. For example, the same person may incidentally find that they make new friends and come to appreciate this social gain.
A distinction can be made between subjective and objective gains. A subjective gain is the person's own perception that they have gained something through participation, e.g., "I have acquired new skills though this martial art." An objective gain, in contrast, is an external observer's assessment that the person has gained something through participation, e.g., "The individual has scored well on this grading of skill." The present study focuses only on subjective gains, measured by self-report. As in research into volunteering (Clary et al., 1998;Davis, Hall, & Meyer, 2003), there will be one set of scales measuring motives and a separate set of scales measuring gains. Each motive and each gain scale would be expected to be homogeneous (unidimensional), but each motive scale would be expected to be distinct from its corresponding gain scale (separate dimensions). The higher-order structure of motives would be expected to be similar to that previously suggested by Ingledew and Markland (2008; see also Ingledew et al., 2009). However, the higher-order structure of gains might differ from the higher-order structure of motives. This is because, in the translation of motives into gains, there will be perturbations arising from, for example, unsought gains.
There are four good reasons for creating a measure that allows one to examine exercise gains alongside motives. First, individuals with a particular motive may be more likely to make a corresponding gain, and individuals experiencing a particular gain may be more likely to develop a corresponding motive. This would manifest as a positive association between motive and corresponding gain. Such positive associations have been found in research into volunteering with charity organizations and other prosocial behavior such as organizational citizenship behavior (Davis et al., 2003;Finkelstein, 2006Finkelstein, , 2008. Second, some motives may be easier or harder than others to convert into corresponding gains (harder to attain or to perceive). This would manifest as a within person mean difference between motive and corresponding gain. To the best of the authors' knowledge, such motive-gain mean differences have not been examined in any literature. Third, individuals with a particular motive may experience different outcomes (such as level of satisfaction) depending on whether they make a corresponding gain. This would manifest as an interactive effect of motive and corresponding gain. Such interactive effects have been found in research into volunteering (Clary et al., 1998). Fourth, even if exercise gains do not moderate the effects of motives, they could have effects in their own right. This would manifest as an additive effect of motive and corresponding gain. Such additive effects have been found in research into volunteering and other prosocial behavior (e.g., Davis et al., 2003;Finkelstein, 2006Finkelstein, , 2007.

Present Study
The aim of the present study was to develop a measure of motives and gains by adding gains scales to the EMI-2 (Markland & Ingledew, 1997). The resulting composite measure would be known as the EMGI. The objectives were to assess the lower-order structure (factor analysis of items) and the higher-order structure (factor analysis of scales), and to examine discrepancies between motives and corresponding gains (within-person gain-motive differences). The effects of motives and gains on exercise-related processes and outcomes (behavioral regulation, exercise amount, satisfaction, and intention) are considered in another article (Ingledew, Markland, & Strömmer, 2014).
The following findings were expected.
1. Motive and gain items would reflect their intended constructs, and motive and gain constructs, though correlated, would be distinct. 2. The higher-order structure of motives would be similar to that identified by Ingledew and Markland (2008), that is to say, health-fitness, appearance-weight, social engagement, and enjoyment related groupings. The higher-order structure of gains might be somewhat different. 3. There would be discrepancies between gains and motives, of varying size and direction. For example, gains that are harder to attain or perceive, such as perhaps health-related gains, would show negative mean differences between motives and gains. Conversely readily attainable or perceptible gains, EXERCISE MOTIVES AND GAINS INVENTORY such as perhaps social gains, would show positive mean differences between motives and gains.

Design and sample
The study was a cross-sectional survey using a questionnaire. Ethical approval was granted by a university departmental ethics committee. Participants were adults between 18 and 35 years of age. They were recruited from communal areas of a British university (e.g., kitchens and lounges of halls of residence, cafeterias, seating areas), rather than from sport and exercise facilities, so as to ensure a wide range of exercise participation levels. A total of 210 individuals completed the questionnaire. However, 14 (7%) of these did not complete the gains section of the questionnaire because they had not engaged in any exercise in the past 12 months. Therefore, the effective sample size was 196. Of these, 60% (118/196) were women and 40% (78/196) men. The mean age was 22.12 years (SD of 3.08). The mean BMI was 22.53 (SD), and 55% belonged to a club in order to participate in sport or recreational physical activity.

Measures
EMGI. The EMGI comprised a motives section and a gains section. The motives section was the EMI-2 (Markland & Ingledew, 1997). The instructions and stem for this section (see Appendix) invited participants to focus on their personal reasons as to why they exercise or might exercise. The items (see Appendix) were of the form "To …," or "Because …," or "For …." The gains section was newly created. The instructions and stem for this section (see Appendix) invited participants to focus on their personal experience of exercise and what they had gained from it. The items (see Appendix) were of the form "I have …," or "I have been able to …," or "It has allowed me to …," or "It has enabled me to …." Each gain item corresponded to a particular motive item. For example, the gain item "[My personal experience of exercise has been that] it has helped me to maintain good health" corresponded to the motive item "[I exercise] to maintain good health." For each gain item, the wording was determined by consensus between the three authors. For both motives and gains sections, response options ranged from "not at all true for me" (0) to "very true for me" (4). The order of items was randomized, separately for each section. Each section was comprised of 51 items forming 14 scales of 3 or 4 items each.
Other measures. The motives and gains measures were presented along with other measures, in the following order: exercise motives (EMGI) and behavioral regulation of exercise (Behavioral Regulation in Exercise Questionnaire 2; Markland & Tobin, 2004); exercise amount, stage of change, and intention; affect (Positive and Negative Affect Scale; Watson, Clark, & Tellegen, 1988); exercise gains (EMGI) and exercise satisfaction. The order of measures was designed to flow well, while separating gains from motives. Analyses using some of the other measures are reported elsewhere (Ingledew et al., 2014).

Analyses
Missing values. Missing values were imputed. As there were only seven missing data points, single imputation by expectation-maximization was used (Olinsky, Chen, & Harlow, 2003).
CFA of items. The EMGI item scores were subjected to CFA using Mplus version 7 (Muthén & Muthén, 2012). This was to assess how well items reflected intended constructs and whether motive and gain constructs were distinct. A CFA approach was adopted in this study because the gain items were developed based on existing motive items from the EMI-2. The 14-factor structure of the EMI-2 has been previously established with items reflecting their intended constructs and constructs being discriminated from each other (Markland & Ingledew, 1997). Due to the well-established factor structure of the EMI-2, there were strong hypotheses for the factor structure of the gain scales. Because of these hypotheses, the CFA approach was deemed appropriate. Based on the factor structure of the motives scale, a series of 14 two-factor models was tested. In each model, a motive construct was examined alongside the corresponding gain construct. To illustrate, in the twofactor model for affiliation, the four motive items were free to load onto one factor, the four gain items onto another factor ( Figure 1). The affiliation motive and affiliation gain factors were allowed to correlate. The measurement errors of corresponding motive and gain items (e.g., "To make new friends" and "I have made new friends") were also allowed to correlate, to accommodate their matching content. For each model, the Satorra-Bentler scaled χ 2 was examined (Satorra & Bentler, 1994), which adjusts for multivariate non-normality, the comparative fit index (CFI), and the standardized root mean square residual (SRMR). Following Hu and Bentler (1999), the criterion for adequate fit was a combination of CFI close to .95 and SRMR close to .08. For completeness, the authors also report the root mean square error of approximation (RMSEA). The discriminant validity of the scales was assessed by calculating the average variance extracted (AVE) (Fornell & Larcker, 1981) and comparing that to the inter-scale correlations. Discriminant validity is considered to be confirmed when the AVE estimates for both constructs (motive and gain) are greater than their shared variance (i.e., square of the correlation) (Fornell & Larcker, 1981). Having established the factor structures of the items, the motive and gain scale scores were then computed as the means of item scores (i.e., unit weighted composite scores).
Exploratory structural equation modeling (ESEM) of scales. The EMGI scale scores were subjected to ESEM following procedures outlined by Asparouhov and Muthén (2009) within Mplus version 7 (Muthén & Muthén, 2012). This was to explore the higher-order factor structures of motives and gains. The ESEM approach was deemed appropriate as motives had a hypothesized higher-order factor structure based on previous work on the EMI-2 (Ingledew & Markland, 2008) but gains were not necessarily expected to exhibit the same higher-order factor structure. The ESEM approach was adopted in preference to CFA because it was not hypothesized that the structures of motives and gains would necessarily be the same, nor were strictly simple factor structures hypothesized. ESEM was adopted in preference to traditional exploratory factor analysis because it provides a range of fit statistics, and allows comparison of models to determine the optimal number of factors. The term "exploratory structural equation modeling" is potentially misleading when applied purely to factor analysis as this does not include the estimation of structural relations between latent variables. Nevertheless, ESEM procedures have been widely used for factor analytic purposes (cf., Guay, Morin, Litalien, Valois, & Vallerand, 2015). Guay et al. (2015) drew a distinction between the use of ESEM as a confirmatory factor analytic procedure, where the number of factors are specified a priori, and its use as an exploratory factor analytic procedure where model fit information is used to determine the optimal number of factors to extract. In the current study, ESEM was used in an exploratory fashion. Separate analyses were conducted for motives and gains. For each of these, six models were sequentially fitted to the data, systematically increasing the number of factors from one to six and the models were compared using Satorra-Bentler χ 2 difference tests (ΔSatorra-Bentler χ 2 , Satorra & Bentler, 2001), with alpha set to .01 due to the susceptibility of this approach to lead to over-factoring (cf., Myers, Chase, Pierce, & Martin, 2011). Oblique (promax) rotation was used. In each analysis, the number of factors was constrained, but (in contrast to CFA) each item was free to load on any factor. For ESEM with promax rotation and a robust estimator, Mplus only produces the Satorra-Bentler scaled χ 2 , root mean square residual (RMR), and RMSEA. The authors report these as well as the CFI, calculated by hand.
Differences between means. To compare mean differences between motives and gains, a t-test and correlation were then conducted for each pairing of motive scale with its corresponding gain scale. 1 Negative mean differences denote a lower gain relative to the original motive, conversely positive mean differences denote a higher gain relative to the original motive.

CFAs of Items
The results of the CFAs of item scores are shown in Table 1. The Satorra-Bentler scaled χ 2 was non-significant for 11 of Example two-factor CFA model using affiliation motive and gain.
1 It was not possible to compute the difference between factor means in the CFAs. Multi-group CFA was not appropriate, since motives and gains were not separate groups. In principle, latent change analysis might have been appropriate (motives changing into gains). However, in practice, this would have required strong factorial invariance across motives and gains (same configuration, equal loadings, and intercepts), whereas it was only hypothesized configural factorial invariance (same configuration, unconstrained loadings, and intercepts). the 14 models, though not for enjoyment, social recognition, and weight management. However, all 14 models met the criteria for adequate fit according to the CFI (≥.95) and SRMR (≤.08). Factor loadings were greater than .60 for 92 of the 102 items. The lowest loadings were for the appearance motive item "look younger" (.47) and the health pressures motive item "recover from an illness or injury" (.49). The correlations of motive factors with corresponding gain factors were all positive, and the 95% confidence intervals of these correlations all had lower boundaries above .00 and upper boundaries below 1.00, except for revitalization which had an upper boundary of 1.00 (95% CI [.86, 1.00]). Discriminant validity was satisfactory for affiliation, appearance, competition, ill-health avoidance, nimbleness, positive health, and strength and endurance. The AVEs for the corresponding motives and gains scales were smaller than their shared variance for challenge, enjoyment, health pressures, revitalization, social recognition, and stress management. The AVE for weight management motive (.70) was higher than the shared variance (.55), whereas the AVE for weight management gain was lower (.52).

EXERCISE MOTIVES AND GAINS INVENTORY
The means, standard deviations, internal consistencies, and motive-gain correlations of the scales are shown in Table 2. Means (on a scale from 0-4) were not strikingly low or high, except for health pressures motive (M = 0.96) and gain (.87), and positive health motive (3.23). Cronbach's alpha was above .70, with the exception of health pressures motive (α = .54), health pressures gain (α = .68), and revitalization motive (α = .68). The correlations of motive scales with corresponding gain scales were all significant and positive, and were notably high for enjoyment (r = .83), competition (.84), and stress management (.86).
The factor structures of motives ( Figure 2) and gains ( Figure 3) were similar in many respects. Both motives and gains had a social engagement factor, encompassing affiliation, challenge, competition, and social recognition. Both also had an enjoyment/revitalization factor,   encompassing enjoyment, revitalization, and stress management. Both had a negative health factor, encompassing health pressures and ill-health avoidance. Both also had a health/fitness factor, encompassing positive health, strength/ endurance and nimbleness, and also to some extent illhealth avoidance. However, motives had an appearance/ weight management factor, whereas gains had a weight management factor, with appearance gain loading predominantly on the health/fitness factor. Correlations between factors were more positive for gains (range .15 to .64) than for motives (−.18 to .51).

Differences between Means
The differences between motives and gains composite scores are shown in Table 2. There were significant positive differences (gain greater than motive) for affiliation, challenge, enjoyment, nimbleness, social recognition and strength and endurance, and negative differences (motive greater than gain) for ill-health avoidance and positive health.

Main Findings
The results were consistent with expectations, with minor exceptions. All EMGI items reflected their intended constructs. There were two factor loadings that were comparatively low. These were the appearance motive item "look younger," and the health pressures motive item "recover

62
from an illness or injury." It is likely that these items do not apply well to the present sample of healthy young adults, but could apply to other samples. Correlations between motive factors and corresponding gain factors were all positive, and the 95% confidence intervals excluded 1.00 except for revitalization where it touched 1.00. Discriminant validity was confirmed by AVE for affiliation, appearance, competition, ill-health avoidance, nimbleness, positive health, and strength and endurance. AVE for challenge, enjoyment, health pressures, revitalization, social recognition, stress management, and weight management did not fully support discriminant validity. Discriminant validity for these scales based on the AVE's is an issue, and arises from the similar wording and content of the corresponding motive and gain scales. Further research is required to establish the discriminant validity of the scales. The true test of the discriminant validity of the scales will be whether they have differential predictive capabilities in practice. Some support for this has already been established (Ingledew et al., 2014). The higherorder structures of motives and gains were similar in many respects. However, appearance motive was associated with weight management, whereas appearance gain was associated with health and fitness. Gain factors were more positively intercorrelated than were motive factors. There were significant mean differences between some motives and gains. Positive mean differences suggested a higher degree of attainability for affiliation, challenge, enjoyment, nimbleness, social recognition, and strength and endurance, whereas negative differences suggested a low degree of attainability for ill-health avoidance and positive health, with appearance, competition, health pressures, revitalization, stress management, and weight management. Overall, the aim, to develop a measure of gains that corresponded to the existing EMI-2 measure of motives, was met.

Theoretical Implications
It seems that people can distinguish between their motives and their gains. Even though the motive and gain items were similar in wording and proximal in time, the psychometric results indicate clear separation of constructs for many of the scales used. Results indicated less than desirable discriminant validity for challenge, enjoyment, health pressures, revitalization, social recognition, stress management, and weight management. The motives for enjoyment, revitalization, stress management, and challenge are prominent in habitual exercisers and could have come into an alignment with their corresponding gains due to individuals pursuing activities they know will fulfill their motives. The scales for health pressures, positive health, social recognition, and weight management could represent gains that are difficult for participants to gauge because these gains are slow to materialise and/or difficult to perceive. It remains to be seen whether discriminant validity for these scales might improve when the motives and gains measures are used at a longer time interval. Nevertheless, even if highly correlated, they might be useful for research because asking about motives and gains in this way taps into a natural form of discourse.
Having asked someone what they want from exercise ("I want to lose some weight") it is natural to then ask them what they have gained ("No, I haven't lost much weight, but I have felt much more relaxed"), and perhaps odd not to ask. The positive associations between motives and corresponding gains (evident in the factor-factor correlations and the scale-scale correlations) could reflect two possible causal relationships. It may be that people who strive for something (motive) are generally more likely to attain it and to notice if they do attain it (gain). It may alternatively or additionally be that people come to seek (motive) what they happen to get and appreciate (gain). They may not have initially been aware that such a gain was possible, or that they would appreciate it. Correlations may be particularly high for some motives such as enjoyment and competition because these are prominent in regular exercisers (Ingledew et al., 1998) in whom there would be more opportunity for motives and gains to come into alignment. Whatever the explanation for the positive correlations, previous research findings on the effects of motives may have been confounded by unmeasured but correlated gains. In light of this, previous conclusions about the effects of motives will need to be re-evaluated (see Ingledew et al., 2013).
The differences in higher-order structures of motives and gains are intriguing. From motive to gain, appearance shifted its association from weight management to health and fitness. Perhaps initially people see weight loss as the primary means to improve their appearance. However, in due course they come to recognize that physical changes such as muscle tone and agility also convey a positive impression. These different characteristics of appearance motive and appearance gain may reflect a shift in body image from investment (excessive preoccupation) to evaluation (constructive management; cf., Carraça et al., 2011). From motive to gain, intercorrelations between factors became more positive. Perhaps initially most individuals want a limited number of things out of exercise. However, they subsequently experience a range of other benefits, which they acknowledge as personal gains. The different higher-order structures of motives and gains mean that one would need to think carefully and perhaps do preliminary analyses before aggregating motives and gains when studying their effects.
There may be two reasons for the mean within-person differences between gains and motives. Some gains may be easier to actually attain, or easier to perceive, or both. For example, ill-health avoidance may be difficult to attain (could take a long time) and to perceive (few overt signs). Affiliation may be relative easier to attain (exercise often has a strong social element) and to perceive (plenty of overt signs). Weight management may be difficult to attain in practice but easy to perceive should it happen. However, it is important to note that around each of the mean within-person differences, there was extensive individual difference (SD), so that in all instances there were some individuals who gained less than they wanted, some who gained about what they wanted, and some who gained more than they wanted.

Future Directions
Further research is needed to ascertain the psychometric properties of the EMGI in a variety of populations. From a theoretical perspective, the value of the instrument will lie in the study of the consequences and causes of motives and gains. It will provide a means of studying how motives and gains influence exercise-related processes and outcomes, including their interactive effects (see Ingledew et al., 2013). It will also provide a means of studying how motives and gains arise, including how they influence each other. Longitudinal and experimental designs will allow for more precise determination of causality. Such studies of change over time and response to interventions are particularly illuminating in the ongoing effort to validate the use of an instrument (Messick, 1995). It might be possible to measure gains in other ways such as developing standalone gains measures, but the instrument created here is presented as a means of measuring gains corresponding to an existing measure of motives for exercising. Ultimately, the instrument may merit use within public health programs to promote physical activity.

Motive and Gain Items
The instructions for motives were "Following are a number of statements concerning the reasons people often give when asked why they exercise. Whether you currently exercise regularly or not, please read each statement carefully and indicate, by circling the appropriate number, whether or not each statement is true for you personally, or would be true for you personally if you did exercise…." The stem was "Personally, I exercise (or might exercise)…." The instructions for gains were "This section of the questionnaire can only be completed by people who have some current or recent experience of exercise. So if you have not exercised within the last 12 months, please just put a cross here and skip this section. The questions are about what you have actually gained from exercise. This may be the same or different from what you originally wanted or hoped to gain. Please tell us your personal experience of exercise using the following scale…." The stem was "My personal experience of exercise has been that…."

Concept
Motive item Gain item